$BE : negative news circulating worth being aware of. Short seller report from Hunterbrook Capital and Crossroads Capital alleges Bloom Energy remains heavily dependent on Chinese suppliers for scandium oxide - a critical rare earth mineral used in fuel cell production. This directly contradicts CEO KR Sridhar’s on-record statement that the company had “no China supply chain” for the mineral. The supply concern - scaling to meet AI data centre demand would require approximately 220 tons of scandium oxide annually. Global supply is projected at just 240 tons. Short sellers argue this makes long term manufacturing targets physically unattainable. Securities fraud investigations opened following the report. Stock down ~12% in a single session. Bloom’s response - filed an official 8-K with the SEC branding the allegations as “false and misleading.” But here’s the context most people are missing. The $NBIS x $BE deal wasn’t a forward thinking clean energy partnership. It was a desperate pivot after Plan A collapsed. Nebius’s original US data centre strategy relied on standard gas turbines. Local communities and environmental regulators blocked them - noise, carbon emissions, air quality permits. Plan A was legally dead. Meanwhile Nebius had promised Wall Street $7-9B ARR by end of 2026. Missing that because they couldn’t turn the lights on would have decimated the stock. Bloom Energy offered the only viable escape route - behind the meter fuel cells requiring no combustion permits, promising 328MW commissioned rapidly within the same year. Nebius accepted significantly higher long term power costs to protect their short term timeline. Think of it as paying overnight shipping rates because the deadline is unmovable. Now the short seller report questions whether Bloom can actually deliver on that promise. Why this matters for $NBIS - three execution risks emerge simultaneously: 1/ Timeline bottleneck - Bloom manufacturing backlogs delay fuel cell delivery. Nebius cannot power up data centres on schedule. The early 2027 Vera Rubin delivery window for Meta is under direct threat. 2/ $META relationship strain - Meta already has “Meta Compute” ambitions. A delayed Nebius gives them justification to pull back on the optional $15B compute expansion. 3/ Capital strain - Nebius has raised its 2026 annual capex guidance to $20-25B with deeply negative free cash flow of -$3.68B reported in 2025 and accelerating. Any forced pivot to a third alternative power source adds significant capital pressure at the worst possible time. And this is exactly why $IREN’s vertical integration model matters. $IREN owns the grid connections directly. For the Childress Texas campus powering the $MSFT contract - IREN connected directly into ERCOT, built their own mega-substations and manages the transformers on site. No fuel cell manufacturer. No scandium oxide supply chain. No third party counterparty risk. $NBIS - at key US sites, paid premium costs for speed and accepted third party dependency where grid queues made traditional connection impossible. $IREN - paid premium costs upfront to own the infrastructure outright. The mitigating factor - Bloom filed an 8-K branding the allegations as false and misleading. If their defence holds the risk dissipates. But the $BE situation just exposed the structural vulnerability at the heart of the neocloud asset light model. Power. Chips. Construction. Capital. All four need to land perfectly. On time. Simultaneously. $BE $NBIS $IREN NFA